Idle. Not listening.
Recognition only listens for the language(s) checked here — this replaces blind auto-detection. Pick just one for the most reliable results. Picking more than one keeps a self-correcting switch between just your checked languages, but the sentence where you actually switch languages may still come out garbled (a real limitation of browser speech recognition — see docs).
⚠️ Experimental — off by default. Guesses when the speaker changed from pauses + voice volume (not real voiceprint/speaker recognition). Works best when two people speak at clearly different volumes; if speakers sound similar it will misattribute lines. See known limitations in the project docs.
Web Speech API (SpeechRecognition) is required. Try Chrome on desktop or Android. Safari/iOS support is limited or missing — see the project's known limitations.
Nothing said yet. Tap the mic button above and start talking.